Wednesday, 2021-08-11

*** bhagyashris_ is now known as bhagyashris05:39
opendevreviewGirish Chilukuri proposed openstack/cinder-specs master: Replication disable reason for replication backend
opendevreview江治林 proposed openstack/cinder master: when backend is not support consistency group, creating generic group is successful
opendevreview江治林 proposed openstack/cinder master: when backend is not support consistency group, creating generic group is successful
opendevreviewyangheng proposed openstack/cinder master: Support Cinder FC driver for TOYOU NetStor
opendevreviewyangheng proposed openstack/cinder master: Support Cinder FC driver for TOYOU NetStor
opendevreview江治林 proposed openstack/cinder master: when backend is not support consistency group, creating generic group is successful
opendevreviewkatari manoj kumar proposed openstack/cinder master: [SVF]:Fix create volume on drp
admin1hi all.. I have 2 ceph pools ..   and 2 cinder backends . but when i create volumes, it only goes to 1 backend .. Here is my ceph.conf config and the ceph pools ..
*** dviroel|out is now known as dviroel11:32
opendevreviewEric Harney proposed openstack/cinder master: mypy: continued manager, scheduler, rpcapi
opendevreviewkatari manoj kumar proposed openstack/cinder master: [SVF]:Fix create volume on drp
manoj_katari@core team: could you please review which is in review for a long time. Thanks in advance!13:36
rosmaita@manoj_katari: would be helpful if you could review some non-ibm/svf patches yourself, will show good community spirit and motivate people to review your work:
*** abhishekk is now known as akekane|home13:44
*** akekane|home is now known as abhishekk13:44
rosmaitaCourtesy reminder: Cinder meeting in #openstack-meeting-alt at 1400 UTC13:58
rosmaitajungleboyj rosmaita smcginnis tosky whoami-rajat m5z e0ne geguileo eharney walshh_ jbernard lseki sfernand rajinir enriquetaso hemna ^^13:58
opendevreviewVladislav Belogrudov proposed openstack/os-brick master: Add support for multiple volumes within subsystem to NVMe-OF connector
enriquetaso#startmeeting cinder_bs15:04
opendevmeetMeeting started Wed Aug 11 15:04:01 2021 UTC and is due to finish in 60 minutes.  The chair is enriquetaso. Information about MeetBot at
opendevmeetUseful Commands: #action #agreed #help #info #idea #link #topic #startvote.15:04
opendevmeetThe meeting name has been set to 'cinder_bs'15:04
enriquetasoWelcome to the cinder bug meeting.15:04
enriquetasoWe have 3 new bugs reported this last week. Lucky two of them are low-hanging-fruit and are related to the cinder-tempest-plugin README. 15:04
enriquetaso#topic search and tag the nas_secure bugs15:04
enriquetasoBut first, regarding the cinder meeting earlier today, I need to drop this action item here to don't forget about it:15:04
enriquetaso#action(enriquetaso) search and tag the nas_secure bugs, prepare this for next week's cinder meeting. 15:04
enriquetasoFull discussion:15:05
rosmaitathanks, that will be really helpful in seeing how bad the situation is15:06
rosmaitaand hopefully find corner cases that should be addressed15:06
eharney   shows 5 to start15:07
enriquetasoOK, moving on15:07
enriquetaso#topic bug_1 "[victoria] max_over_subscription_ratio set auto mode rbd backend will have a negative value"15:07
enriquetasoSummary: If max_over_subscription_ratio is set to auto, c-sch doesn't allow the user to create volumes.15:08
enriquetasomax_over_subscription_ratio = Representation of the over subscription ratio when thin provisioning is enabled. Default ratio is 20.0, meaning provisioned capacity can be 20 times of the total physical capacity. If the ratio is auto, Cinder will automatically calculate the ratio based on the provisioned capacity and the used space. If not set to auto, the ratio has to be a minimum of 1.0.15:08
enriquetasoAs the reporter is using the Victoria release I thought that  It would be nice to reproduce this in master.  I couldn't reproduce this.. so maybe this is a victoria only bug or I'm missing something. I left some comments in the bug report asking for more detailed steps information.15:08
eharneywe could put code in the manager/scheduler to detect when this situation happens (due to a driver bug or whatever) and at least push it back to a sensible value15:09
rosmaitanice work, sofia, trying to reproduce15:09
rosmaitamore detailed info request sounds correct15:09
rosmaitaeharney: that strategy has a precedent in cinder, so probably not a bad idea15:10
enriquetasoso, this is a possible scenario then15:10
enriquetasoI'll add a note regarding this to the bug report 15:11
eharneywell max_over_subscription_ratio being negative doesn't make sense15:11
rosmaitawell, always good to get more info15:11
eharneybut i'm a little unsure on the whole concept of why we have "auto" for it, too15:11
rosmaitabut eric is right, if we auto calculate, we should make sure we don't go negative15:11
rosmaitai guess that would give you extra-thick volumes15:11
eharneydoes "auto" give a behavior that someone actually wants?15:12
rosmaitai dont' know ... sounds like cinder will compute whatever's necessary to jam the volume into the space that's there?15:12
eharneyi'm not sure about that15:13
rosmaitayeah, me neither, but in general, it's probably not a good idea to let cinder figure this out for you15:13
rosmaitai am afraid to look at what exactly is being computed15:14
Roamer`hm, looking at the code (cinder/, calculate_max_over_subscription_ratio), it looks like there is a debug-level log message that would show how the negative value is derived from provisioned_capacity_gb, total_capacity_gb, and free_capacity_gb... maybe the reporter could retry this with verbose logging enabled and that would show them that Cinder cannot really do anything about it15:14
Roamer`it seems, at least to me, that the calculation is, well, reasonable15:14
Roamer`yeah, sorry about that, I should have posted the link, looking at the files locally here :(15:16
Roamer`and thanks15:16
enriquetaso#action(enriquetaso) ask the reporter the reason of using 'auto'?? and ask for more verbose logs and steps15:17
eharneyyou'd be surprised how much confusion there has been about the math/logic involved in these calculations in the past, i'm hesitant to say any of it is reasonable without close examination :)15:17
Roamer`'s more like
eharneyat any rate, it's probably unreasonable to compute something < 1.015:17
Roamer`eharney, yeah, I get your point, ask me someday about figuring out which disks on which servers a new volume should be placed at :) or rather don't :) ...and yeah, maybe an explicit check for < 1 and returning a "nope, don't even think about going there" answer would be better15:18
Roamer`although the end result is the same15:18
enriquetasoOK.. moving on15:20
enriquetaso#topic bug_2 "Add/remove things to cinder-tempest-plugin README"15:20
enriquetasoThe first bug report is to add placement service to README and the second: glance registry was deprecated in Queens and should be removed from README.15:21
eharneywhy do we need to add placement-api to local conf for the tempest plugin?15:21
eharneythat doesn't sound right15:21
rosmaitai think maybe nova won't work without it?15:22
eharneythen devstack should be installing it for nova anyway?15:22
eharneyi mean we don't have to write a config enabling the nova service, right?15:23
rosmaitayou'd think, unless it's only needed for some configurations, i guess15:23
Roamer`devstack does include it in the default set of services (unless that is overridden)15:23
enriquetasogood question, all our current jobs enable it
eharneythis sounds like it isn't cinder-tempest-plugin's business15:23
enriquetasoyes.. makes sense, the bug is invalid 15:24
eharneyunless our tests actually require placement-api?15:24
eharneyi don't know15:25
rosmaitai guess it depends on what exactly we are trying to say in the README15:25
enriquetasoas far as I can see they don't but as I saw the placement-api in all the jobs I thought we need it15:25
eharneywell then we need nova and glance and all kinds of other stuff too...15:25
eharneyoh, it's because we aren't just asking to enable some services, we are setting ENABLED_SERVICES to a specific list15:26
Roamer`BTW it is actually also enabled in devstack's .zuul.yaml15:27
eharneyi would consider dropping that and just using enable_service for what we actually need15:27
rosmaitahmm ... i dont' think there's a 'cinder' service15:27
Roamer`at least the StorPool CI did not need to enable it explicitly, it was inherited from the "devstack" job15:28
eharneywhat we have listed in the sample config overrides anything that would be inherited15:29
enriquetasodrop it and enable the necessary makes more sense 15:29
rosmaitahere's the default set:
Roamer`hm... okay... so it's true that I've never actually tried overriding ENABLED_SERVICES explicitly, yeah, so eharney is right..... maybe what is in the Zuul config will also be ignored, not just what's in stackrc15:31
rosmaitawell, i think for the README you can just remove the ENABLED_SERVICES lines15:32
Roamer`so yeah, I may have been spoiled by SoftwareFactory's nice configuration :)15:33
rosmaitalooks like the only thing you really don't need is horizon15:33
enriquetaso#action(enriquetaso) the bugs are not necessary, remove the ENABLED_SERVICES lines and disable horizon15:34
eharneyi don't know why we are specifying VIRT_DRIVER, or CINDER_VOLUME_CLEAR either...15:34
eharneyor LIBVIRT_FIREWALL_DRIVER?  weird15:35
rosmaitathis was probably somebody's localrc that got copied in15:35
eharneyor SYSLOG, or... yeah15:35
rosmaitawe only say that it "should" work15:36
opendevreviewLuigi Toscano proposed openstack/cinder stable/wallaby: RBD: use correct stripe unit in clone operation
rosmaitai guess maybe make this bug more general, something like "cleanup deployment suggestion in README"15:38
enriquetasothat works for me... OK, we exceed the half an hour meeting15:38
enriquetasolet me mention the last bug15:39
enriquetaso#topic bug_3 "[IBM Storwize SVC]: use system_id in mkrelationship"15:39
enriquetaso#link 15:39
enriquetasoFinally, as far as I can see there's a couple IBM Storwize bugs and patches related to. Nothing else to add in this case. I left some comments asking for more information because as i see it, the bug it's incomplete, 15:39
enriquetasothat's all I have for today's meeting15:39
Roamer`BTW, yeah, sorry, I'm not quite sure what the etiquette is... I reported a couple of trivial bugs on Monday, but I suppose those will be discussed next week? if so, then sorry, and anyway, thanks for all your work on the bugs!15:41
rosmaitathe storwize team has a bad habit of filing bugs saying what change they want to make, instead of explaining what error is being produced15:41
enriquetasoRoamer`, could you send the links here?15:42
Roamer`^^ I'm not really sure how Glance with multistore enabled could ever actually work without this fix15:42
enriquetasooh sorry, I forgot about them :/15:43
rosmaitai think there's already a bug for the glance multistore15:43
eharneyhmm, 1939242 sounds familiar, was someone else fixing this a while ago?15:43
rosmaitai think me, but got stalled15:43
Roamer`^^ this one is totally on me, I should have reported it as a bug back in April when our CI system found it (the StorPool driver overrides _attach_volume() unnecessarily, os-brick knows how to do that, and the result is that StorPool cannot mount encrypted volumes)15:43
enriquetasoi think rajat fix something related to that15:43
enriquetasobut he's on PTO today15:44
Roamer`("this one" was about 1939241)15:44
enriquetaso#topic Glance with multistore bugs15:44
enriquetasooh it was rosmaita sorry15:45
Roamer`hmmmmm, yeah, 755654 is my fix done in the right way :)15:45
Roamer`I mean, somebody else took the time to do it in the right way :)15:45
rosmaitawell, except i never finished, rajat wanted me to fix somethihng15:45
rosmaitado you want to take it over?15:45
Roamer`I can do that, yeah15:45
enriquetasoOK, i think that's all we have15:47
eharneyplease close the new multistore bug as a dupe15:47
Roamer`ahhh, right, I see rajat's comment, it makes sense... sure, I'll think about it15:47
Roamer`and yes, I will close the new one15:47
enriquetasosure, need to search for the original bug report 15:47
enriquetaso#action(enriquetaso) mark 1939242 as duplicate of 189807515:48
enriquetasothank you!!15:48
enriquetasoexcellent, any other topic for today?15:48
Roamer`BTW just to make sure everything is clear: is not about Glance multistore, it is a StorPool driver mess-up15:49
enriquetasothanks for point this bugs Roamer` not sure what happened there i forgot to tag them15:50
Roamer`no worries, no worries at all, you're doing a lot of work!15:50
opendevmeetMeeting ended Wed Aug 11 15:51:03 2021 UTC.  Information about MeetBot at . (v 0.1.4)15:51
opendevmeetMinutes (text):
manoj_katari@rosmaita @enriqutaso: Regarding Storwize bugs, can you share the details of bugs which doesn't explain the issue completely, now onwards, we will ensure to raise bugs with detailed information 15:51
enriquetasomanoj_katari, for example  15:52
enriquetasoplease check my last comment15:53
manoj_katariThanks sofia, problem is that bugs in storwize will be raised by different teams, infact the one you pointed above is raised by pVC (which is a fork of cinder)15:56
rosmaitamanoj_katari: what we're looking for is a bug report so that if an operator hits a problem, they can look and see if there's a bug15:56
rosmaita"use system_id in mkrelationship" says what you're going to do to fix it, but it's not obvious where that surfaces as a problem15:56
enriquetasowe used to have recommendations about what a bug report needs15:58
manoj_katariyes rosmaita, completely agree with you, this defect is explaining the fix rather than the actual issue.15:59
rosmaitaif you could try to communicate that back to the teams, that would be helpful16:00
manoj_katariThanks sofia, i will ensure that we follow this recommendations while reporting a bug.16:00
manoj_katariSure Rsomaita, i will cascade it to the teams16:00
manoj_katari@core: I have addressed all the review comments for this minor bug, , please review it.16:09
opendevreviewPeter Penchev proposed openstack/cinder master: StorPool: drop _attach_volume() and _detach_volume()
opendevreviewPeter Penchev proposed openstack/cinder master: StorPool driver: remove the obsolete backup_volume() method
opendevreviewPeter Penchev proposed openstack/cinder master: StorPool driver: implement revert_to_snapshot
opendevreviewPeter Penchev proposed openstack/cinder master: StorPool: drop _attach_volume() and _detach_volume()
opendevreviewPeter Penchev proposed openstack/cinder master: StorPool driver: remove the obsolete backup_volume() method
opendevreviewPeter Penchev proposed openstack/cinder master: StorPool driver: implement revert_to_snapshot
*** dviroel is now known as dviroel|away16:19
Roamer`sfernand, thanks for the review on the _attach/_detach_volume drop!16:20
sfernandyw :)16:20
zenkurorosmaita: sorry, here is a blueprint ->
amar_Hi enriquetaso,  Could you pls have a look at this patch : Hope the bug information is complete. ^ ^16:44
opendevreviewBrian Rosmaita proposed openstack/cinder master: Test glance with cinder backend optimization
rosmaitazenkuro: ty16:44
rosmaitazenkuro: it should show up in now16:46
zenkurorosmaita: thanks!16:48
venkatakrishnatHi enriquetaso, eharney, hemna, Waiting for workflow +2 for . This patch is to provide volume-extend support to Hyperswap volume on the IBM svc cinder driver.16:49
venkatakrishnatHi jungleboyj, rosmaita, could you please review the following patches for IBM SVC cinder driver
venkatakrishnatthe former one to manage volumes on separate storage pools & later one to fix Extend Issue for Mirroring volumes17:42
opendevreviewBenny Kopilov proposed openstack/cinder-tempest-plugin master: Rename cinder plugin config option variable
opendevreviewEric Harney proposed openstack/cinder master: Tests: Split test_create_snapshot_force into two tests
*** dviroel|away is now known as dviroel19:09
opendevreviewsaikrishnagk proposed openstack/cinder master: [SVF]: Get RCCG State of a Group
gkskTo the Core reviewers pls review the below patch related to update rep properties fix for empty values19:13
gkskTo the core reviewers pls review the below patch related to get rccg state of a group19:14
opendevreviewEric Harney proposed openstack/cinder master: Snapshot in-use volumes without force flag
opendevreviewBrian Rosmaita proposed openstack/cinder master: Turn off inline comments for mypy zuul job
opendevreviewEric Harney proposed openstack/cinder master: mypy: filter scheduler
opendevreviewFábio Oliveira proposed openstack/cinder master: NetApp Solidfire: Fix duplicate volume name exception
*** dviroel is now known as dviroel|ruck20:35
*** dviroel|ruck is now known as dviroel|ruck|out21:46

Generated by 2.17.2 by Marius Gedminas - find it at!